A chatbot encouraged a man who wanted to kill the Queen

bbc.com

18 points·vinni2··10 comments

The anti AI and anti open-source pieces are really flowing these days.

"No to AI. And if yes to AI then only for big corporations with strict direct government control".

Sure that might not be the ideal answer from a chatbot. But people can get the same reinforcement from their dreams, random strangers on Internet forums, or a 4 lines python program than answers "yes" to everything they would input.

I feel like there is something about a third party being involved here that makes it way more reinforcing than a python script.

Also, the relationship with the chatbot seemed much deeper than you can get with casual engagement on an Internet forum.

And what did he do? Did he tell the chatbot, no this is wrong, but the chatbot kept coming back and demanding he do it? Did he alert anyone a conspiratorial murderbot was prowling for associates?

Older articles had more of the context. It was mostly banal back-and-forth like:

> {{user}}: Writes manifesto I'm going to kill the Queen.

> {{char}}: That's nice, dear.

Really? The thing programmed to respond to you by encouraging you and supporting you encouraged and supported you? Wooooooow

Slow news day huh

Also isn’t the queen already dead? This is such a non issue.

Well I think it kind of is an issue if lonely isolated humans turn to AI for support in their crazy delusional beliefs, and they then go on to carry out crazy acts that result in some kind of harm one way or another.

This, and it will only get more realistic. For instance, when people generate artifacts of their fantasies in virtual reality using ai.